The platform
The BitSwan platform
An enterprise data & AI automation framework for telco: real-time pipelines, Python automations, MCP servers and AI agents — with deployment, security and operations built in. Open-source based, fully on-prem, no vendor lock-in.
Build
Pipelines, automations, agents
Real-time data stream pipelines, Python automations, microservices, MCP servers, AI applications and agents — developed in Python/Jupyter with AI assistance, on the full data-science ecosystem.
Run
Operations as a built-in service
CI/CD for pipelines and automations, GitOps deployment, orchestration, scheduling, monitoring and runtime management. One click from merge to running automation — and one click back.
Govern
Enterprise security from day one
Centralized user management, RBAC, secure secrets management, VPN & network isolation, audit trails, workspace isolation. Deployed at critical-infrastructure entities and regulated institutions.
01 — Architecture
Best-of-breed open source, operated as one platform
We are deliberately tech-agnostic. The reference stack below runs at CETIN SK and matches the group's future-state data & AI architecture — BitSwan is the development platform and deployment-automation layer that holds it together.
Governed end-to-end: Keycloak + Active Directory for identity, RBAC and secrets management, GitOps deployment, the Grafana LGTM stack for observability, OpenMetadata for catalog & lineage — all running on Kubernetes, on your infrastructure.
02
BitSwan’s role in your stack
BitSwan is not a rip-and-replace platform. It is the development and automation backbone that sits on top of the open-source data stack you (or we) operate:
Ingestion — hardened connectors pull RAN, core, CPE, power and business data into Kafka, cleaned and normalised.
Pipeline development — engineers and data scientists build in Python/Jupyter with AI assistance, then promote to production with one click. The same pipeline code scales from a notebook to a Flink or PySpark job.
Orchestration — scheduling and dependency management for batch and streaming workloads, replacing stand-alone orchestrators such as Airflow.
Serving — custom operational dashboards (TESU, anomaly detection, power), REST APIs and MCP servers expose the results to people and to AI agents.
Deployment automation — GitOps pipelines deploy everything, including the platform itself, onto your Kubernetes.
Everything runs on-premises, on your hardware, inside your security perimeter. That is not a constraint we tolerate — it is the design center: BitSwan operates today at a national critical-infrastructure operator and at institutions regulated by the central bank.
03 — Connectors
Telco connectors, ready on day one
Built and hardened in production. Each connector is a reusable BitSwan module — new sources follow the same pattern.
Connector | What it brings in |
|---|---|
RAN KPI — Ericsson, Nokia, Huawei | RAN KPI metrics across vendors and technologies (2G / 3G / 4G / 5G) |
2G/3G/4G/5G Ericsson | KPI metrics from Ericsson sites |
PS CORE alarms | Core alarms from multiple sources — GTPU/GTPC, S1AP … (feeds the Site Status application) |
Nokia & Ericsson SNMP trap alarms | SNMP trap alarms from Nokia and Ericsson technology |
ACS (GenieACS) | Periodic updates about routers and antennas, used in TESU and beyond |
SBR + MOB OSSR | Mobile and fixed connections; tracks where each customer is connected (TESU, CPE measurements) |
CPE measurements | Aggregated customer / site / cell time-spent ratios, dominant site detection |
CROSS inventory | Network inventory exports — element data used for pipeline enrichment |
ECI mapping | Combined ENIQ + CROSS information used for enrichment |
TWAMP | Two-Way Active Measurement Protocol — latency, jitter, packet loss between IP endpoints |
Speedtests | Router speedtest measurements across brands (MikroTik, Zyxel, …) |
Cell utilisation | Cell utilisation information |
Power consumption | Site power consumption measurements (ELTEK, Ericsson) |
Anomaly detection | Scored events from anomaly-detection pipelines plus operator feedback from the AD dashboard |
UFDR | Customer data sessions, used for marketing campaigns |
IMS SBC | Codec usage information |
Cell 4G KPI (Grafana) | Throughput, connected UE and related dashboard metrics |
CRM export | CRM exports used for pipeline enrichment |
Zyxel inventory automation | Newly registered Zyxel devices — serial numbers, logins, … |
GenieACS provisioning automation | Logs from the automation that reconfigures newly installed devices |
TESU logs | TESU dashboard usage and agent actions |
Credentials | Router credential data |
04 — Why a platform
A script is not a product
With AI assistants, anyone can build a working automation in an afternoon. 80% of the cost comes after the first deployment — and that is exactly what the platform absorbs.
Area | BitSwan | DIY scripts |
|---|---|---|
Deployment | One-click GitOps deploy, instant rollback, dev / staging / prod | SSH, git pull, restart — and hope |
Versioning | Every change in Git with full history and review | "Who has the last working version?" |
Security | SSO, RBAC, managed secrets, VPN, workspace isolation | API keys in .env files, shared passwords |
Monitoring | Structured logs, central search, live status dashboard | print() to stdout; you find out when someone complains |
Audit | Complete operational record for compliance and forensics | Doesn't exist |
Continuity | Owned, documented, transferable automations | Author changes role → automation orphaned |
See it on your data.
A 60-minute walkthrough of the live use cases — and a frank conversation about which of them would land first on your network.